American Society for Mass Spectrometry Conference 2013

1,573 views

Published on

Published in: Technology, Education
0 Comments
0 Likes
Statistics
Notes
  • Be the first to comment

  • Be the first to like this

No Downloads
Views
Total views
1,573
On SlideShare
0
From Embeds
0
Number of Embeds
5
Actions
Shares
0
Downloads
0
Comments
0
Likes
0
Embeds 0
No embeds

No notes for slide
  • 165 studies for 75 different principal investigators for 12,589 samples, a roughly doubled number compared to the budget year before the WCMC was implemented (and about 4-fold higher than in the 2010/11 budget year).  we have 8 LCMS 3 Agilent QTOF, 1 Leco Citius HRTOF, 1 ABSciex QTRAP, 1 ABSciex TripleTOF,   1 Thermo LTQ, 1 Thermo LTQ-FTICR,  and 7 GCMS 4 Leco TOF, 2 Agilent quad, 1 Agilent QTOF
  • American Society for Mass Spectrometry Conference 2013

    1. 1. Multivariate and network tools for analysis and visualization of metabolomic dataDmitry Grapov, Oliver FiehnWest Coast Metabolomics Center, Genome CenterUniversity of California Davis
    2. 2. State of the art facility producing massiveamounts of biological data…>13,000 samples/yr>160 studies~32,000 data points/study(80 samples x 400 metabolites)
    3. 3. Stylized Analysis at the Metabolomic ScaleNucl. Acids Res. (2008) 36 (suppl 2): W423-W426.doi: 10.1093/nar/gkn282
    4. 4. Analysis at the Metabolomic Scale
    5. 5. Challenges ofMetabolicAnalysis• Complex W I D E data ( variables >> samples)• Inefficient access to domain knowledge• Difficulty of translating experimental findings intoactionable biological interpretation(s)• Biological• Analytical
    6. 6. Cycle of Scientific DiscoveryData AcquisitionDataNetwork MappingHypothesis GenerationData ProcessingHypothesis
    7. 7. Network Mapping1. Statistical andmultivariatemodels2. Network ofmetaboliterelationships3. Mappinganalysis results tothe network
    8. 8. Can you spot the difference?Univariate Multivariate Predictive ModelingANOVA PCA PLS
    9. 9. Predictive Modeling• PLS, OPLS, NNMF, random forests, ANN, SVM, etc.Clustering• HCA, k-NN, k-means, SOM, etc.change in timedifference between groups1. Multivariate Modelingcase vs. control x treatmenttime-dependent differencesamong groups
    10. 10. 2. Network GenerationDefine connections between metabolitesBiochemical (substrate/product)•Database lookup•Web queryChemical (structural orspectral similarity )•fingerprint generationEmpirical (dependency)•correlation, partial-correlationBMC Bioinformatics 2012, 13:99 doi:10.1186/1471-2105-13-99
    11. 11. 3. Network MappingAnalysis results Network Annotation Mapped Network
    12. 12. Treatment Effects Network=MetabolitesShape = increase/decreaseSize = importance (loading)Color = correlationConnectionsred = Biochemical relationshipsviolet = Structural similarity
    13. 13. Time Course Network
    14. 14. Gaussian Markov Network (intervention)Partial correlationNetwork
    15. 15. Available Tools (free as in speech and beer)http://sourceforge.net/https://github.com/dBioinformatics. 2012 Sep 1;28(17):2288-90. doi: 10.1093/bioinformatics/bts439Biological Database Translations in RChemical Translation System (CTSgetR)Chemical Identifier Resolver (CIRgetR)https://github.com/
    16. 16. dgrapov@ucdavis.edumetabolomics.ucdavis.eduThis research was supported in part by NIH 1 U24 DK097154

    ×